Multi-dimensional perception method and system for power transmission line state based on multi-source data fusion

Abstract

The application relates to the technical field of power transmission lines, and discloses a power transmission line state multi-dimensional sensing method and system based on multi-source data fusion. Through a fault hierarchical judgment mechanism, the fault evolution is divided into the budding period, the development period and the critical period, and a differentiated caching strategy is matched in each stage to realize the collaborative optimization of fault data retention and resource utilization. Through a full-link value quantification model, three dimensions of fault state, time decay and data type are fused, and confidence gain, link adaptation and redundancy loss correction are provided to provide a unified priority standard for caching, routing and transmission, and improve the transmission efficiency of the communication window. A bimodal reconstruction strategy is adopted, the fault physical mechanism is deduced and the space-time correlation interpolation is combined, missing data is reliably completed through fusion judgment and conflict analysis, and the reconstruction result conforms to the physical law, so that complete time sequence support is provided for fault tracing.

Description

Technical Field

[0001] This invention relates to the field of power transmission line technology, and in particular to a method and system for multi-dimensional sensing of power transmission line status based on multi-source data fusion. Background Technology

[0002] Existing power transmission line condition monitoring systems typically rely on public communication networks to transmit data collected on-site, such as voltage, current, temperature, and vibration, to a remote monitoring center. For transmission lines deployed in remote mountainous areas, many towers are located in areas without operator communication signals. Traditional monitoring devices cannot achieve real-time data transmission in these areas, creating "perception blind spots," forcing maintenance personnel to rely on manual inspections to obtain line status information. While existing technologies attempt to extend communication coverage through wireless ad hoc networks and relay nodes, they suffer from bottlenecks such as limited link coverage, high node deployment costs, and poor data transmission reliability in complex mountainous terrain.

[0003] When transmission lines experience transient faults such as voltage dips or current surges, the interruption of communication links prevents the timely transmission of key fault characteristic data to the monitoring center. This leads to a significant delay in fault location and cause analysis, forcing maintenance personnel to rely solely on manual inspections after the fault occurs, resulting in low response efficiency. Simultaneously, a large amount of monitoring data is discarded or overwritten during communication interruptions, hindering fault tracing due to a lack of complete time-series data support and making it difficult to accurately reconstruct the entire fault evolution process. Ensuring the complete retention of transmission line status sensing data and the priority transmission of critical information in communication-dead zones without relying on public network communication has become a core challenge restricting the widespread application of multi-dimensional transmission line status sensing technology in remote mountainous areas. Summary of the Invention

[0004] The technical problem to be solved by this invention is that the existing technology has the disadvantage of interrupted data transmission of state perception in remote mountainous areas where the monitoring device for power transmission lines is in a communication blind zone. To address this, we propose a multi-dimensional state perception method and system for power transmission lines based on multi-source data fusion.

[0005] To achieve the above objectives, this application adopts the following technical solution: a multi-dimensional sensing method for transmission line status based on multi-source data fusion, comprising:

[0006] Collect multi-source heterogeneous status data on the side of the transmission line tower and generate a status data package with timestamps;

[0007] Local data caching modules are deployed on towers in communication blind spots. The caching strategy is dynamically adjusted according to the communication link status. When the communication quality is lower than the preset threshold, the caching priority of transient fault characteristic data is increased. Multi-level caching queues are used to store steady-state data and transient event data respectively.

[0008] Along the communication blind zone, select aggregation nodes to deploy wireless ad hoc network communication modules and construct a hierarchical tree-like ad hoc network topology. When the tower outputs a suspected fault status judgment, carry a status flag bit and the comprehensive value score V of the whole link data in the header of the data packet. Adjacent nodes forward the data according to the comprehensive value score.

[0009] During the satellite communication window, the aggregation node prioritizes and transmits data packets according to the comprehensive value score V of the entire data link.

[0010] The remote monitoring center uses a time window sliding alignment algorithm based on time tags to splice data packets from different interruption periods along the time axis. For data missing sections, a dual-modal reconstruction strategy combining fault physical mechanism and spatiotemporal correlation is adopted to output the continuous status trajectory of each monitoring point along the transmission line and the complete time sequence of fault evolution during the communication interruption.

[0011] Furthermore, the local data caching module constructs a lightweight edge inference unit based on a physical feature library of typical transmission line faults on the tower. When the real-time data collected by the sensor shows an evolution trend consistent with any fault mode in the physical feature library for three consecutive sampling periods in time, the lightweight edge inference unit autonomously determines that the corresponding tower has entered a suspected fault state without communicating with the remote monitoring center. The suspected fault state determination result is embedded as metadata into the header of all subsequent cached data packets, and all subsequent data collected by the corresponding tower is forcibly promoted to the highest priority cache queue.

[0012] Furthermore, the lightweight edge inference unit and the local data caching module construct a fault hierarchical judgment-caching strategy closed-loop linkage mechanism. The mechanism divides the fault evolution process into multiple progressive stages, and each stage is matched with differentiated cache control logic.

[0013] In the initial stage of a fault, the priority of the first-level cache is increased for the matched fault feature dimension data; in the stage of fault development, multiple copies of the fault feature data are redundantly stored to lock the full amount of data before and after the fault determination for a preset period.

[0014] During the worsening phase of the fault, a dynamic preemption mechanism for cache resources is triggered, pausing the cache update of non-core steady-state data and allocating most of the local cache space to the full fault data.

[0015] The mechanism has a built-in reverse verification logic for misjudgments. It performs reverse matching verification on the identified fault characteristics at a preset period. If the fault characteristics continuously fall back to the normal threshold range, the cache priority promotion strategy is automatically lifted.

[0016] Furthermore, a full-link value quantification model for fault-related data of transmission lines in communication blind spots is constructed. The model is based on the fault correlation degree, time sensitivity, and physical importance of data packets, and is superimposed with nonlinear gain of fault matching confidence, link communication state adaptation correction, and data retention reliability loss correction. The comprehensive value score V of the full-link data is obtained through a three-level progressive composite calculation.

[0017] The first level is based on three dimensions: fault correlation degree, time sensitivity, and data physical importance. The basic value benchmark value of the data packet is calculated by linear weighting through preset weight coefficients. The fault state weight distinguishes between three states: normal, suspected fault, and fault criticality. The time decay factor reflects the proximity between the data acquisition time and the fault determination time. The data type weight reflects the difference between steady-state data and transient data.

[0018] The second level introduces a nonlinear gain correction for fault confidence. Based on the confidence level of fault feature matching, the baseline value is nonlinearly amplified, so that high-confidence fault data and low-confidence data have a significant difference in comprehensive value score, ensuring that key data has priority in resource competition.

[0019] The third level performs full-scenario adaptation correction, adjusts data priority in communication blind spot environments through link state adaptation factors, and reduces the transmission value of multiple replica data through redundancy loss correction items to avoid redundant data occupying scarce transmission resources, and finally outputs a unified full-link data comprehensive value score.

[0020] Furthermore, when the lightweight edge inference unit of a certain tower outputs a suspected fault status judgment, the corresponding tower carries a status flag bit and the full-link data comprehensive value score V in the header of the status data packet, and actively broadcasts the suspected fault status judgment information to neighboring nodes. The neighboring nodes establish temporary high-weight routing entries in their routing tables for the corresponding tower that outputs a suspected fault status. The forwarding weight of the routing entry is positively correlated with the full-link data comprehensive value score V of the data packet. When forwarding data, the status data packets from the corresponding tower are forwarded preferentially from high to low according to the comprehensive value score.

[0021] After receiving a status data packet carrying a suspected fault status flag, the aggregation node establishes a fault tracking record for the corresponding tower in its local storage unit. During the subsequent satellite communication window, the status data packet of the corresponding tower is marked with priority according to the comprehensive value score V of the whole link data. At the same time, a confirmation command is sent in the opposite direction to the corresponding tower, triggering the corresponding tower to package and upload the intermediate evolution data in the process of matching the physical feature database of typical faults of the transmission line.

[0022] Furthermore, the aggregation node constructs a data lifecycle hierarchical model based on the fault evolution stage. The data lifecycle hierarchical model completes the priority ranking of all data packets based on the comprehensive data value score V carried in the header of the status data packet.

[0023] The aggregation node dynamically calculates the maximum amount of data that can be transmitted in the current window based on the preset start and end time and window duration of the satellite communication window, combined with the real-time link bandwidth, and selects status data packets in descending order of the overall data value score V of the entire link to enter the transmission queue.

[0024] Before the satellite communication window ends, the transmission rate is dynamically adjusted according to the remaining window duration. The analysis value of the status data packets that have not yet been transmitted is determined based on the comprehensive value score V of the whole link data before the start of the next satellite communication window. Status data packets with a comprehensive value score lower than the preset threshold, weak fault correlation and exceeding the fault analysis time window are intelligently discarded and the discard log is recorded. Status data packets with a comprehensive value score higher than the preset threshold and strong correlation with the current suspected fault status are forcibly retained and marked as pending transmission even if the time window has expired.

[0025] Furthermore, in the dual-modal reconstruction strategy, the first mode is a deduction model based on the physical mechanism of the fault. It uses multi-source data of the same tower in adjacent time windows before and after the fault determination, combined with the fault type prediction result output by the lightweight edge inference unit, and calls the transmission line physical equation corresponding to the fault type to deduce the theoretical change trajectory of each physical quantity in the missing period.

[0026] The second mode is an interpolation model based on the spatiotemporal correlation of multiple towers. It uses monitoring data of adjacent nodes on the same line within the same time period and in similar fault evolution stages to generate probability estimates of missing data through fault time sequence alignment and spatial correlation analysis.

[0027] Furthermore, the theoretical change trajectory of the first modality output is fused with the probability estimate of the second modality output. When the difference between the two is less than a preset threshold, the weighted average is taken as the reconstructed data. When the difference between the two is greater than the preset threshold, the multimodal conflict resolution mechanism is triggered. Based on the confidence level of the fault evolution stage output of the lightweight edge inference unit, the change pattern of similar working conditions in the historical operation data of the corresponding tower, and the data consistency of adjacent nodes in this period, a comprehensive decision is made to select the modality output with higher confidence or to correct the reconstructed data according to the fault mechanism priority principle.

[0028] Furthermore, based on the high-precision time stamps, node geolocation codes, and fault determination metadata embedded in the lightweight edge inference unit carried by each status data packet, the remote monitoring center uses a time window sliding alignment algorithm to splice status data packets from different aggregation nodes and different interruption periods along the time axis, and marks the fault evolution timeline of each corresponding tower according to the fault determination metadata; for the sections with missing data on the time axis, the dual-modal reconstruction strategy is executed, and finally the continuous status trajectory and complete fault evolution time sequence of each monitoring point along the transmission line during the communication interruption are output.

[0029] The present invention also provides a multi-dimensional sensing system for the status of transmission lines based on multi-source data fusion, the system comprising:

[0030] The tower-side data acquisition and caching unit includes multiple types of sensors, a unified timing module, and a local data caching module. It is used to collect multi-source heterogeneous status data and generate time stamps. In communication blind spots, it dynamically adjusts the caching strategy according to the link status. It also has a built-in lightweight edge inference unit for on-site determination of suspected fault status.

[0031] The local ad hoc network routing unit is deployed at the aggregation node and adjacent towers to build a hierarchical tree-like ad hoc network topology and to perform dynamic priority forwarding based on the tower status flag and the comprehensive value score V of the full-link data.

[0032] The satellite backhaul and scheduling unit, deployed at the aggregation node, includes a satellite communication terminal and a data lifecycle hierarchical scheduling module, which is used to prioritize and hierarchically transmit data packets based on the comprehensive value score V of the entire link data within the satellite window period;

[0033] The remote monitoring center includes a data receiving module, a time window sliding alignment module, a dual-modal reconstruction module, and a fault tracing module. It is used to splice data packets from different interruption periods, perform dual-modal reconstruction on missing data segments, and output continuous state trajectories and fault evolution sequences.

[0034] The technical effects and advantages of this invention are as follows:

[0035] This invention employs a fault stratification and caching strategy closed-loop linkage mechanism to divide the fault evolution process into three progressive stages: the incipient stage, the development stage, and the critical stage. Each stage is matched with differentiated caching control logic. In the incipient stage, only the priority of characteristic data is increased to avoid resource waste. In the development stage, triple-copy redundant storage is activated to lock critical data. In the critical stage, dynamic preemption of cache resources is performed, allocating over 90% of the space to the full fault data. This mechanism solves the core problems of traditional caching strategies, such as the inability to distinguish fault evolution stages during communication interruptions and the easy overwriting of critical data, achieving synergistic optimization of fault data retention rate and cache resource utilization.

[0036] This invention employs a full-link value quantification model for fault-related data in transmission lines within communication blind zones. This model combines three dimensions—fault state weight, time decay factor, and data type importance—with nonlinear gain of fault confidence, link state adaptation correction, and redundancy loss correction in a three-tiered progressive composite calculation. This provides a unified priority determination standard for the entire link, including local caching, local routing, and satellite transmission. This model solves the underlying technical problem of insufficient coordination in distributed priority rules, enabling suspected faulty tower data to receive priority forwarding rights in the local self-organizing network. It achieves accurate and prioritized backhaul of high-value data within the satellite communication window, while a smart discarding mechanism prevents invalid data from occupying scarce transmission resources, significantly improving data transmission efficiency within a limited communication window.

[0037] This invention employs a dual-modal reconstruction strategy based on fault physics and spatiotemporal correlation. The first mode derives the theoretical trajectory of change based on the fault physics equation, while the second mode generates probability estimates based on the spatiotemporal correlation of multiple towers. Through dual-modal fusion judgment and conflict resolution mechanisms, reliable completion of missing state data is achieved during communication interruption periods. This technique solves the problem that traditional data completion methods rely solely on statistical interpolation and lack physical mechanism constraints, potentially leading to reconstruction results that violate electrical operating laws. It ensures that the reconstructed data not only conforms to the physical essence of the fault but also fully utilizes spatial correlation information, providing complete time-series data support for fault tracing. Attached Figure Description

[0038] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:

[0039] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0040] Figure 2 This is a logic diagram of the fault stratification and caching strategy closed-loop linkage mechanism of the present invention.

[0041] Figure 3 This is a flowchart of the dual-modal state trajectory reconstruction and fault tracing process of the present invention;

[0042] Figure 4 This is the overall system flowchart of the present invention. Detailed Implementation

[0043] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0044] See Figure 1 As shown, this invention provides a multi-dimensional sensing method for the status of transmission lines based on multi-source data fusion. This method aims to address the problem of interrupted data transmission for status sensing from transmission line monitoring devices in remote mountainous areas within communication blind zones. As a preferred embodiment, this method specifically includes the following steps:

[0045] Step 1: Multi-source heterogeneous state data acquisition and fault feature-driven differentiated local caching.

[0046] In this embodiment, multiple types of sensors, including voltage sensors, current sensors, temperature sensors, and vibration sensors, are deployed on the transmission line tower side to synchronously collect multi-source heterogeneous state data at a preset frequency. The multi-source heterogeneous state data refers to the collective term for electrical and physical parameters collected by different types of sensors during the operation of the transmission line, whose data structures and physical meanings differ.

[0047] Specifically, during the data acquisition process of each sensor, a unified timing module completes end-to-end clock synchronization. This module outputs a nanosecond-level synchronization clock signal based on the BeiDou Navigation Satellite System, providing a unified time reference for all sensing devices and generating a high-precision time tag for each set of acquired data. The high-precision time tag, also called a timestamp, has nanosecond-level time accuracy. Its core function is to achieve time-series synchronization and full lifecycle traceability of multi-source data, ultimately forming a status data packet with a timestamp. The status data packet is a standardized data encapsulation unit, integrating three core types of information within a single sampling period: multi-source monitoring data, a high-precision time tag, and a unique identifier for the acquisition node.

[0048] Furthermore, local data caching modules with communication link status awareness are deployed on the corresponding towers in communication blind spots. These local data caching modules are embedded functional units deployed on the transmission line towers in communication blind spots, possessing three core capabilities: communication link status awareness, hierarchical data storage, and dynamic adjustment of caching strategies. This module monitors the communication link quality with adjacent nodes or aggregation nodes in real time and dynamically adjusts the caching strategy based on the link status. Adjacent nodes are upstream and downstream transmission line tower monitoring nodes with direct wireless communication links to the target tower; aggregation nodes are core monitoring nodes of transmission line towers with wireless self-organizing network relay and satellite communication backhaul capabilities; communication link quality is a quantitative indicator characterizing the reliability of wireless communication link transmission between tower nodes and adjacent nodes or aggregation nodes, corely covering three key parameters: signal-to-noise ratio, bit error rate, and link connectivity; the caching strategy is a set of caching priorities, storage durations, redundancy, and eviction rules set for different types of monitoring data.

[0049] As a preferred solution, when communication quality falls below a preset threshold, the local data caching module proactively prioritizes caching data for transient fault characteristics such as voltage dips and current surges. "Transient" refers to a short-term, non-steady-state process where the system transitions from one steady-state operating state to another, typically lasting from milliseconds to seconds (e.g., voltage spikes caused by lightning strikes, current surges during short circuits, electromagnetic oscillations after circuit breaker operation). This contrasts with "steady-state" (stable parameters during normal operation). A multi-level cache queue is used to store steady-state data and transient event data separately, with a redundant storage mechanism for transient event data. This mechanism prioritizes the retention and traceability of critical fault data during communication interruptions, preventing the loss of critical transient data due to being overwritten by steady-state data. Among them, transient fault characteristic data are characteristic data that can characterize the transient process before and after a transmission line fault occurs, and the core includes non-steady-state operating data such as voltage dips, current surges, and short-term vibration shocks; multi-level buffer queues are multi-level data storage queues divided according to data priority, with different queues corresponding to different buffer retention durations, storage redundancy, and transmission priorities; steady-state data are electrical and physical parameter data with smooth changes and no abrupt changes collected when the transmission line is in a normal and stable operating state; transient event data are the full monitoring data corresponding to events in which the operating state of the transmission line changes abruptly and triggers the fault characteristic identification threshold; the redundant storage mechanism is a multi-copy backup storage mechanism used for high-priority data, the core function of which is to avoid data loss during communication interruptions and ensure the integrity of fault tracing data.

[0050] Meanwhile, in this embodiment, the local data caching module constructs a lightweight edge inference unit based on a physical feature library of typical transmission line faults on the corresponding tower. The physical feature library of typical transmission line faults is a pre-built standardized database that comprehensively covers the temporal evolution patterns of physical features, fault criteria, and feature thresholds for typical fault types such as electrical short circuits, open lines, icing, and tower tilting. The lightweight edge inference unit is a low-power inference computing unit deployed within the local caching module on the tower side. It can achieve on-site fault state identification based on the physical feature library of typical transmission line faults, without relying on cloud computing resources.

[0051] More preferably, when the real-time data collected by the sensors shows an evolution trend consistent with any fault mode in the physical feature library of typical faults in transmission lines for three consecutive sampling periods, the lightweight edge inference unit autonomously determines that the corresponding tower has entered a suspected fault state without communicating with the remote monitoring center. The unit then embeds the suspected fault state determination result as metadata into the header of all subsequent cached data packets. Simultaneously, all subsequent data collected from the corresponding tower is forcibly elevated to the highest priority cache queue until communication with the corresponding tower is restored or the lightweight edge inference unit determines that the fault evolution trend has been eliminated based on subsequent data. Through this on-site edge determination mechanism, the tower can still autonomously identify fault risks even when communication is interrupted, and the determination result is bound to the data packet, providing a basis for priority decisions in subsequent routing and transmission. The suspected fault state is the operating state of the corresponding section of the transmission line tower, determined on-site by the lightweight edge inference unit after matching the typical fault evolution trend with continuously sampled data; the metadata is data used to describe the core attributes of the monitoring data packet, including four key types of information: fault state identifier, data acquisition node information, timestamp, and priority level.

[0052] See Figure 2As shown, to further enhance the reliability of fault data retention and the accuracy of edge detection in scenarios of complete communication outage, this embodiment constructs a fault hierarchical detection-caching strategy closed-loop linkage mechanism for the lightweight edge inference unit and the local data caching module. This mechanism divides the fault evolution process into three progressive stages: fault initiation, fault development, and fault criticality. Each stage is matched with differentiated cache control logic: In the fault initiation stage, corresponding to scenarios where the data matches the fault feature library by more than 60% but does not meet the judgment threshold for three consecutive cycles, only the matched fault feature dimension data is prioritized by one level, while retaining the basic cache quota for all steady-state data to avoid cache resource occupation caused by non-fault mutations; In the fault development stage, corresponding to scenarios where fault patterns are matched for three consecutive sampling cycles and a suspected fault state judgment is triggered, the highest priority cache forced promotion strategy of the original scheme is executed, and three-copy redundant storage of fault feature data is started to lock the fault judgment. The full data of the 10 sampling periods before and after the fault is determined will not be discarded. During the fault critical period, which corresponds to the scenario where the fault characteristics continue to intensify for more than 5 consecutive sampling periods and the matching degree exceeds 90%, the dynamic preemption mechanism of cache resources is triggered. The cache update of non-core steady-state data is suspended, and more than 90% of the local cache space is allocated to the full fault data. At the same time, the intermediate feature data of the fault evolution process is fully retained to provide complete time-series support for subsequent fault tracing. The full fault data refers to all original monitoring data and intermediate feature data related to the fault evolution from the preset period (such as 10 sampling periods) before the fault determination time to the period of fault feature elimination, including transient and steady-state data of all dimensions such as voltage, current, temperature, and vibration.

[0053] Meanwhile, this closed-loop linkage mechanism incorporates a reverse verification logic for misjudgments. Every two sampling cycles, it performs a reverse matching verification on the identified fault characteristics. If the fault characteristics fall back to the normal threshold range for two consecutive cycles, the cache priority enhancement strategy is automatically lifted, releasing cache resources and preventing cache resource exhaustion and transmission bandwidth waste caused by misjudgments. Through this layered linkage mechanism, adaptive allocation of cache resources at different fault stages is achieved. While ensuring the complete retention of fault data, it effectively avoids resource waste in non-fault scenarios. The reverse verification logic for misjudgments refers to re-matching and verifying the fault characteristics identified as "suspected fault status" at a preset period (e.g., every two sampling cycles). If the matching degree falls back below the normal threshold range for two consecutive cycles, the cache priority enhancement strategy is automatically lifted, preventing cache resource waste and transmission bandwidth occupation caused by misjudgments due to momentary interference.

[0054] To achieve a unified priority determination standard across the entire link of local caching, local routing, and satellite transmission, and to address the core underlying problem of insufficient coordination in the original decentralized priority rules, this embodiment adopts the FDFVQ model (Fault Data Full-link Value Quantification) for fault-related data in transmission lines in communication blind spots. FDFVQ provides core quantitative support for the priority management of data across the entire link. The comprehensive value score V of the entire link data output by this model is the sole quantitative basis for determining the data priority of all subsequent stages. Figure 1 With appendix Figure 4 The score V in the model refers to the overall value score of the entire data link. Specifically, this model uses the fault correlation degree, time sensitivity, and physical importance of data packets as the three core dimensions, superimposed with nonlinear gain of fault matching confidence, link communication state adaptation correction, and data retention reliability loss correction. The overall value score of the data packet is obtained through a three-level progressive composite calculation. The higher the overall value score, the higher the priority of the data link. Through this unified quantitative model, the local cache, local routing, and satellite transmission links adopt a consistent priority judgment standard, eliminating the priority conflict and low coordination efficiency caused by decentralized rules. The data retention reliability loss correction refers to the reduction of the overall value score through a redundancy loss correction term for data with multiple replicas. Since there are duplicates among the replicas, sending all of them indiscriminately during transmission would waste bandwidth. The lower the score, the less likely multiple replicas of data will occupy scarce transmission resources at the same time.

[0055] The first level of the model calculates the basic fault association value by linearly weighting and fusing scores from three core dimensions to obtain a baseline value for the basic data packet. Specifically, the fault state dimension is represented by a fault state weight factor S, where S is 1 for a fault-free state, 4 for a suspected fault state, and 8 for a critical fault state; the time association dimension is quantified by a time decay factor, calculated using the following formula: In the formula, Δt is the time difference between data acquisition and the initial fault determination (unit: seconds), and λ is the fault type time decay coefficient, which is 0.08 for fast transient faults and 0.02 for slow evolution faults. Slow evolution faults refer to fault types whose fault characteristics change slowly over time, such as conductor icing, insulator contamination accumulation, and slow tower tilting. Their evolution time is often measured in minutes, hours, or even days, in contrast to fast transient faults (milliseconds to seconds). The data type dimension is represented by the physical importance weight D of the data type. D is 1 for steady-state environmental data, 3 for conventional electrical steady-state data, 6 for transient fault characteristic data, and 10 for intermediate fault evolution characteristic data. The three dimensions are linearly weighted by preset weight coefficients α, β, and γ, which satisfy α + β + γ = 1. The default values ​​are α = 0.5, β = 0.3, and γ = 0.2, which can be dynamically adjusted according to maintenance needs.

[0056] The second level of the model is a nonlinear gain correction for fault confidence. This correction factor amplifies the difference in the overall value score of high-confidence fault data. The calculation formula is as follows: In the formula, M is the fault feature matching confidence level (range [0, 1]), and μ is the fault confidence gain coefficient, with a default value of 2. The base value benchmark is multiplied by the gain correction coefficient to obtain the corrected comprehensive value score. This correction term creates a significant difference in comprehensive value scores between high-confidence fault data and low-confidence data, ensuring that key data obtains absolute priority in resource competition.

[0057] The third level of the model is full-scenario adaptation correction, which is finally adjusted by sequentially using the link state adaptation factor L and the redundancy loss correction term. L is 1 when communication is normal, and 1.5 when communication is in a dead zone. The calculation formula for the redundancy loss correction term is as follows: In the formula, δ is the data redundancy loss coefficient, where δ is 0 for a single copy of data, 0.3 for two copies of data, and 0.5 for three or more copies of data. The corrected comprehensive value score is then multiplied by L and R in sequence to obtain the final comprehensive value score V for the entire link data. Through link state adaptation, data in communication blind spots automatically obtains higher transmission priority; through redundancy loss correction, resource waste caused by multiple copies of data simultaneously occupying transmission bandwidth is avoided.

[0058] Understandably, this model has a clear three-level progressive calculation logic, strong parameter configurability, and can be directly embedded into edge computing units, routing modules, and satellite transmission terminals for stable execution, providing a unified quantitative standard for end-to-end data priority management.

[0059] Step 2: Local data aggregation and routing based on fault perception and tower status weights.

[0060] In this embodiment, solar-powered towers along the communication blind zone are selected as aggregation nodes, and wireless ad hoc network communication modules are deployed there. These wireless ad hoc network communication modules are low-power wireless communication units deployed on the transmission line towers, supporting multi-hop relay between adjacent nodes, dynamic topology adjustment, and high-speed data forwarding.

[0061] Preferably, a hierarchical tree-like self-organizing network topology is constructed between adjacent towers based on the linear geographical topology of the transmission line towers, with the aggregation node as the root node. A dynamic routing strategy is adopted based on the tower number and geographical location, combined with the output status of the lightweight edge inference units of each tower. Specifically, the hierarchical tree-like self-organizing network topology is a multi-level tree-shaped wireless communication network topology constructed according to a linear geographical distribution, with the aggregation node as the top-level root node and adjacent towers along the line as branch nodes; the dynamic routing strategy is a routing planning mechanism that can dynamically adjust the data forwarding path according to the network link quality and node operating status.

[0062] Furthermore, static routing determines the default path for data flow based on the tower sequence number, while dynamic routing makes local corrections to the default path based on the real-time communication link quality. Static routing is a fixed data forwarding path pre-configured based on the linear topology of the transmission line towers.

[0063] More preferably, when the lightweight edge inference unit of a certain tower outputs a suspected fault state determination, the corresponding tower carries a status flag bit and the end-to-end data comprehensive value score V output by the FDFVQ model in the header of the status data packet, and actively broadcasts the aforementioned suspected fault state determination information to neighboring nodes. The neighboring nodes establish temporary high-weight routing entries in their routing tables for the corresponding towers that output suspected fault states. Here, the "routing table" refers to the logical data structure inside a network node (such as the wireless ad hoc network communication module on the tower side) used to store data forwarding paths. The forwarding weight of the routing entries is related to the end-to-end data value of the data packet. Based on the positive correlation of the comprehensive value score V, status data packets from the corresponding towers are prioritized for forwarding according to their comprehensive value scores from high to low during data forwarding. Upon receiving a status data packet carrying a suspected fault status flag, the aggregation node establishes a fault tracking record for the corresponding tower in its local storage unit. During subsequent satellite communication windows, the status data packets from the corresponding tower are prioritized according to the full-link data comprehensive value score V, and a confirmation command is sent in the opposite direction to the corresponding tower, triggering the tower to package and upload intermediate evolution data from the matching process of the transmission line typical fault physical feature database. This routing mechanism ensures that data from suspected faulty towers receives priority forwarding within the local area network, avoiding delays or loss of critical data due to multi-hop transmission. The fault tracking record is a full lifecycle tracking ledger used to record the faulty tower number, fault determination time, and fault data reception progress; the satellite communication window is the time interval during which the satellite communication terminal establishes a stable communication link with the overpassing communication satellite, enabling bidirectional data transmission.

[0064] Step 3: Fault-sensitive data lifecycle hierarchical transmission mechanism for satellite communication windows.

[0065] In this embodiment, the aggregation node integrates a satellite communication terminal and a local storage unit, automatically activating the communication link during the satellite communication window. The satellite communication terminal is a communication terminal device deployed on the aggregation node side, primarily supporting satellite communication link establishment and bidirectional data transmission.

[0066] Step 31: The aggregation node constructs a data lifecycle hierarchical model based on the fault evolution stage. This data lifecycle hierarchical model is a quantitative evaluation model that prioritizes data transmission based on the data fault correlation value and the fault evolution stage. Its core quantitative basis is the end-to-end data comprehensive value score V output by the FDFVQ model. Specifically, this data lifecycle hierarchical model prioritizes all data packets based on the end-to-end data comprehensive value score V carried in the header of the status data packet.

[0067] Step 32: Based on the preset start and end times and duration of the satellite communication window, and combined with the real-time link bandwidth, the aggregation node dynamically calculates the maximum amount of data that can be transmitted in the current window. Status data packets are then selected and entered into the transmission queue in descending order of the overall data value score V. The transmission queue is a sequence of status data packets to be transmitted back via the satellite communication link, ordered by transmission priority. This overall value score ranking mechanism ensures that, within the limited and discontinuous satellite communication window, data with the highest fault correlation and timeliness are given priority for transmission.

[0068] Step 33: Before the satellite communication window ends, the transmission rate is dynamically adjusted based on the remaining window duration. For untransmitted status data packets, their analytical value before the start of the next satellite communication window is determined based on the overall data value score V. Status data packets with an overall value score below a preset threshold, weak fault correlation, and exceeding the fault analysis timeframe are intelligently discarded and their discard logs are recorded. Status data packets with an overall value score above the preset threshold and strong correlation with the current suspected fault state are forcibly retained and marked as pending transmission, even if they exceed the timeframe, ensuring the integrity of fault tracing. This hierarchical scheduling and intelligent discarding mechanism avoids invalid data occupying scarce satellite transmission resources while ensuring the continuous retention of high-value data across window periods. The fault analysis timeframe is the effective retention period for data used to ensure fault analysis and tracing of transmission lines. The fault analysis timeframe refers to the time interval from the moment the fault occurs until the fault data still has reference value for fault tracing and cause analysis. Beyond this window, low-value data can be intelligently discarded to save storage and transmission resources.

[0069] See Figure 3 As shown, step 4: reconstruction of dual-modal state trajectory and fault source tracing based on fault physical mechanism and spatiotemporal correlation.

[0070] In this embodiment, after receiving status data packets from multiple aggregation nodes, the remote monitoring center performs status trajectory reconstruction and fault tracing. The remote monitoring center is a back-end management system deployed within the power grid dispatching agency, primarily responsible for monitoring the operational status of the entire transmission line, fault analysis, and dispatching and handling.

[0071] Step 41: Based on the high-precision time stamps, node geolocation codes, and fault determination metadata embedded in the lightweight edge inference unit carried by each state data packet, a time window sliding alignment algorithm is used to splice state data packets from different aggregation nodes and different interruption periods along the time axis, and the fault evolution timeline of each corresponding tower is marked according to the fault determination metadata. The time window sliding alignment algorithm is a standardized algorithm for time synchronization and alignment of multi-source, multi-node time-series data based on a fixed-length sliding time window; the fault evolution timeline is a time axis marker used to characterize the full-time evolution process of a transmission line fault from its inception and development to its steady state.

[0072] Step 42: For segments with missing data on the timeline, a dual-mode reconstruction strategy combining fault physics mechanism and spatiotemporal correlation is adopted. Fault physics mechanism refers to the fundamental physical laws of electromagnetics, thermodynamics, and materials mechanics that govern the occurrence and development of transmission line faults; spatiotemporal correlation refers to the correlation patterns between the operational data of monitoring nodes at different spatial locations along the transmission line under the same time sequence.

[0073] Specifically, the first mode is a deductive model based on the physical mechanism of the fault. It uses multi-source data such as voltage, current, temperature, and vibration of the same corresponding tower in adjacent time windows before and after the fault determination, combined with the fault type prediction results output by the lightweight edge inference unit, and calls the transmission line physical equations corresponding to the aforementioned fault types to derive the theoretical change trajectory of each physical quantity in the missing time period. The transmission line physical equations are standardized physical equations used to describe the changing laws of the electrical and mechanical operating states of transmission lines. The core equations include thermal balance equations and electromagnetic transient equations for electrical faults, and conductor mechanics equations and vibration mode equations for mechanical faults. The thermal balance equations describe the physical equations that balance heat generation and dissipation in the current-carrying conductors of the transmission line when current flows through them. In this invention, these equations are used to deduce the theoretical trajectory of conductor temperature changes based on current and ambient temperature. The conductor mechanics equations describe the static and dynamic equilibrium equations of the conductors under tension and external loads. In this invention, these equations are used to infer the trajectory of the conductor's stress state based on vibration amplitude, deformation, and other data. The vibration mode equations describe the vibration patterns and natural frequencies of the transmission conductors and fittings at specific frequencies. In this invention, these equations are used to deduce the theoretical vibration modes of mechanical faults such as conductor galloping and aerobatic vibration. This modal analysis fully utilizes the deterministic laws of the physical nature of faults to ensure that the reconstruction results conform to the basic constraints of electrical and mechanical operation.

[0074] The second mode is an interpolation model based on the spatiotemporal correlation of multiple towers. It utilizes monitoring data from adjacent nodes on the same line within the same time period and at similar fault evolution stages to generate probability estimates for missing data through fault sequence alignment and spatial correlation analysis. This mode fully leverages the spatial correlation patterns of monitoring data along the line, using complete data from adjacent nodes to fill in the missing node data.

[0075] Step 43: The theoretical change trajectory output by the first modality is fused with the probability estimate output by the second modality. When the difference between the two is less than a preset threshold, the weighted average is taken as the reconstructed data. When the difference is greater than the preset threshold, the multimodal conflict resolution mechanism is triggered. Based on the confidence level of the fault evolution stage output by the lightweight edge inference unit, the change pattern under similar working conditions in the historical operation data of the corresponding tower, and the data consistency of adjacent nodes during this period, a comprehensive decision is made. The modality output with higher confidence is selected or the reconstructed data is corrected according to the fault mechanism priority principle. Finally, the continuous state trajectory and complete fault evolution time sequence of each monitoring point along the transmission line during the communication interruption are output. Through the above dual-modal fusion and conflict resolution mechanism, reliable completion of missing state data is achieved during the communication interruption period, providing complete time-series data support for fault tracing. Among them, the reconstructed data is the transmission line state data during the communication interruption period obtained by the dual-modal reconstruction strategy; the multimodal conflict resolution mechanism is a processing mechanism that comprehensively decides the optimal reconstruction result through multi-dimensional indicators when there is a significant difference between the two modal reconstruction results.

[0076] It should be noted that the multimodal conflict resolution mechanism refers to the logical processing of selecting or modifying the final reconstruction result when the difference between the reconstruction results of the first mode (physical mechanism deduction) and the second mode (spatiotemporal correlation interpolation) exceeds a preset threshold. This is done by comprehensively considering indicators such as the fault confidence output by the lightweight edge inference unit, the change pattern under similar historical working conditions of the tower, and the consistency of data of adjacent nodes, and by using weighted adjudication or mechanism priority principles.

[0077] In summary, this solution systematically solves the problem of interrupted data transmission for state perception in remote mountainous areas by using local differentiated caching to ensure data retention, unified quantization model to optimize transmission decisions, and dual-modal reconstruction to complete missing trajectories, without relying on public network communication. This significantly improves the applicability and reliability of multi-dimensional state perception technology for power transmission lines in remote areas.

[0078] See Figure 4 As shown, the present invention also provides another technical solution: specifically, it provides a multi-dimensional sensing system for transmission line status based on multi-source data fusion, used to implement the above-mentioned multi-dimensional sensing method for transmission line status based on multi-source data fusion, including:

[0079] The tower-side data acquisition and caching unit, deployed on each tower of the transmission line, includes multiple types of sensors, a unified timing module, and a local data caching module. The multiple types of sensors include voltage sensors, current sensors, temperature sensors, and vibration sensors, used to synchronously acquire multi-source heterogeneous state data at a preset frequency. The unified timing module outputs a nanosecond-level synchronous clock signal based on the BeiDou satellite navigation system, providing a unified time reference for all sensing devices along the line and generating high-precision time tags for each set of acquired data. The local data caching module is deployed on towers corresponding to communication blind spots, possessing communication link status awareness capabilities. It monitors the communication link quality with adjacent nodes or aggregation nodes in real time and dynamically adjusts the caching strategy based on the link status. When the communication quality falls below a preset threshold, it proactively increases the caching priority of transient fault characteristic data. A multi-level caching queue is used to store steady-state data and transient event data separately, with a redundant storage mechanism for transient event data. The local data caching module also incorporates a lightweight edge inference unit based on a typical transmission line fault physical feature library. This unit autonomously determines whether the corresponding tower has entered a suspected fault state when matching fault patterns during continuous sampling periods and embeds the determination result into the data packet header.

[0080] The local ad hoc network routing unit, deployed along the communication blind zone at aggregation nodes and adjacent towers, includes a wireless ad hoc network communication module. Based on the linear geographical topology of the transmission line towers, a hierarchical tree-like ad hoc network topology is constructed between adjacent towers, with the aggregation node as the root node. A dynamic routing strategy is adopted based on tower number and geographical location, combined with the output status of each tower's lightweight edge inference unit. When a tower outputs a suspected fault status, the corresponding tower carries a status flag and a comprehensive data value score V in the header of the data packet. Adjacent nodes create a temporary high-weight routing entry for this tower in their routing tables, and prioritize forwarding the tower's status data packets according to the comprehensive value score from high to low.

[0081] The satellite backhaul and scheduling unit, deployed at the aggregation node, includes a satellite communication terminal, a local storage unit, and a data lifecycle hierarchical scheduling module. The satellite communication terminal automatically activates the communication link during the satellite communication window. The data lifecycle hierarchical scheduling module constructs a data lifecycle hierarchical model based on the end-to-end data comprehensive value score V output by the FDFVQ model. It prioritizes all data packets according to the comprehensive value score, selecting packets in descending order of comprehensive value score to enter the transmission queue. Before the end of the window, it determines the retention value of untransmitted data based on the comprehensive value score, intelligently discarding low-value data and forcibly retaining high-value data and marking it for later transmission.

[0082] The remote monitoring center, deployed within the power grid dispatching agency, includes a data receiving module, a time window sliding alignment module, a dual-modal reconstruction module, and a fault tracing module. The data receiving module receives status data packets from multiple aggregation nodes; the time window sliding alignment module splices data packets from different aggregation nodes and different interruption periods along a timeline and marks the fault evolution timeline for each tower; the dual-modal reconstruction module includes a deductive model based on the physical mechanism of the fault and an interpolation model based on the spatiotemporal correlation of multiple towers, used for dual-modal reconstruction and fusion judgment of data missing sections; the fault tracing module outputs the continuous status trajectory and complete fault evolution time sequence of each monitoring point along the transmission line during the communication interruption period.

[0083] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for multi-dimensional perception of a power transmission line state based on multi-source data fusion, characterized in that, include: Collect multi-source heterogeneous status data on the side of the transmission line tower and generate a status data package with timestamps; Local data caching modules are deployed on towers in communication blind spots. The caching strategy is dynamically adjusted according to the communication link status. When the communication quality is lower than the preset threshold, the caching priority of transient fault characteristic data is increased. Multi-level caching queues are used to store steady-state data and transient event data respectively. Along the communication blind zone, select aggregation nodes to deploy wireless ad hoc network communication modules and construct a hierarchical tree-like ad hoc network topology. When the tower outputs a suspected fault status judgment, carry a status flag bit and the comprehensive value score V of the whole link data in the header of the data packet. Adjacent nodes forward the data according to the comprehensive value score. The local data caching module constructs a lightweight edge inference unit based on a physical feature library of typical transmission line faults on the tower. The lightweight edge inference unit constructs a full-link value quantification model for transmission line fault association data in communication blind spots. The model is based on the fault association degree, time sensitivity, and physical importance of data packets as the basic dimensions, and superimposed with nonlinear gain of fault matching confidence, link communication state adaptation correction, and data retention reliability loss correction. The comprehensive value score V of the full-link data is obtained through a three-level progressive composite calculation. During the satellite communication window, the aggregation node prioritizes and transmits data packets according to the comprehensive value score V of the entire data link. The remote monitoring center uses a time window sliding alignment algorithm based on time tags to splice data packets from different interruption periods along the time axis. For data missing sections, a dual-modal reconstruction strategy combining fault physical mechanism and spatiotemporal correlation is adopted to output the continuous status trajectory of each monitoring point along the transmission line and the complete time sequence of fault evolution during the communication interruption.

2. The method of claim 1, wherein, When the real-time data collected by the sensor shows an evolution trend consistent with any fault mode in the fault physical feature library for three consecutive sampling periods in time, the lightweight edge inference unit autonomously determines that the corresponding tower has entered a suspected fault state without communicating with the remote monitoring center, and embeds the suspected fault state determination result as metadata into the header of all subsequent cached data packets, while forcibly promoting all subsequent data collected by the corresponding tower to the highest priority cache queue.

3. The method of claim 2, wherein, The lightweight edge inference unit and the local data caching module construct a fault hierarchical judgment-caching strategy closed-loop linkage mechanism. The mechanism divides the fault evolution process into multiple progressive stages, and each stage is matched with differentiated cache control logic. In the initial stage of a fault, the priority of the first-level cache is increased for the matched fault feature dimension data; During the fault development stage, multiple copies of the fault characteristic data are redundantly stored to lock the full amount of data before and after the fault determination for a preset period. During the worsening phase of the fault, a dynamic preemption mechanism for cache resources is triggered, pausing the cache update of non-core steady-state data and allocating most of the local cache space to the full fault data. The mechanism has a built-in reverse verification logic for misjudgments. It performs reverse matching verification on the identified fault characteristics at a preset period. If the fault characteristics continuously fall back to the normal threshold range, the cache priority promotion strategy is automatically lifted.

4. The method of claim 2, wherein, The first level of the three-tier system is based on three dimensions: fault correlation degree, time sensitivity, and data physical importance. It calculates the basic value benchmark of the data packet by linear weighting through preset weight coefficients. The fault state weight distinguishes between three states: normal, suspected fault, and fault criticality. The time decay factor reflects the proximity between the data acquisition time and the fault determination time. The data type weight reflects the difference between steady-state data and transient data. The second level introduces a nonlinear gain correction for fault confidence, which nonlinearly amplifies the baseline value based on the confidence level of fault feature matching, so that high-confidence fault data and low-confidence data form a significant difference in comprehensive value score, ensuring that key data has priority in resource competition. The third level performs full-scenario adaptation correction, adjusts data priority in communication blind spot environments through link state adaptation factors, and reduces the transmission value of multiple replica data through redundancy loss correction items to avoid redundant data occupying scarce transmission resources, and finally outputs a unified full-link data comprehensive value score.

5. The method of claim 4, wherein, When the lightweight edge inference unit of a certain tower outputs a suspected fault status judgment, the corresponding tower carries a status flag bit and the full-link data comprehensive value score V in the header of the status data packet, and actively broadcasts the suspected fault status judgment information to neighboring nodes. The neighboring nodes establish a temporary high-weight route entry in the routing table for the corresponding tower that outputs a suspected fault status. The forwarding weight of the route entry is positively correlated with the full-link data comprehensive value score V of the data packet. When forwarding data, the status data packets from the corresponding tower are forwarded first according to the comprehensive value score from high to low. After receiving a status data packet carrying a suspected fault status flag, the aggregation node establishes a fault tracking record for the corresponding tower in its local storage unit. During the subsequent satellite communication window, the status data packet of the corresponding tower is marked with priority according to the comprehensive value score V of the whole link data. At the same time, a confirmation command is sent in the opposite direction to the corresponding tower, triggering the corresponding tower to package and upload the intermediate evolution data in the process of matching the physical feature database of typical faults of the transmission line.

6. The method of claim 4, wherein, The aggregation node constructs a data lifecycle hierarchical model based on the fault evolution stage. The data lifecycle hierarchical model completes the priority ranking of all data packets based on the comprehensive data value score V carried in the header of the status data packet. The aggregation node dynamically calculates the maximum amount of data that can be transmitted in the current window based on the preset start and end time and window duration of the satellite communication window, combined with the real-time link bandwidth, and selects status data packets in descending order of the overall data value score V of the entire link to enter the transmission queue. Before the satellite communication window ends, the transmission rate is dynamically adjusted according to the remaining window duration. The analysis value of the status data packets that have not yet been transmitted is determined based on the comprehensive value score V of the whole link data before the start of the next satellite communication window. Status data packets with a comprehensive value score lower than the preset threshold, weak fault correlation and exceeding the fault analysis time window are intelligently discarded and the discard log is recorded. Status data packets with a comprehensive value score higher than the preset threshold and strong correlation with the current suspected fault status are forcibly retained and marked as pending transmission even if the time window has expired.

7. The method of claim 1, wherein, In the dual-modal reconstruction strategy, the first mode is a deduction model based on the physical mechanism of the fault. It uses multi-source data of the same tower in adjacent time windows before and after the fault determination, combined with the fault type prediction result output by the lightweight edge inference unit, and calls the transmission line physical equation corresponding to the fault type to deduce the theoretical change trajectory of each physical quantity in the missing period. The second mode is an interpolation model based on the spatiotemporal correlation of multiple towers. It uses monitoring data of adjacent nodes on the same line within the same time period and in similar fault evolution stages to generate probability estimates of missing data through fault time sequence alignment and spatial correlation analysis.

8. The method of claim 7, wherein, The theoretical change trajectory of the first modality output is fused with the probability estimate of the second modality output. When the difference between the two is less than a preset threshold, the weighted average is taken as the reconstructed data. When the difference between the two is greater than the preset threshold, the multimodal conflict resolution mechanism is triggered. Based on the confidence level of the fault evolution stage output of the lightweight edge inference unit, the change pattern of similar working conditions in the historical operation data of the corresponding tower, and the data consistency of adjacent nodes in this period, a comprehensive decision is made to select the modality output with higher confidence or correct the reconstructed data according to the fault mechanism priority principle. 9.The power line condition multi-dimensional perception method based on multi-source data fusion of claim 1, wherein, Based on the high-precision time stamps, node geolocation codes, and fault determination metadata embedded in the lightweight edge inference unit carried by each status data packet, the remote monitoring center uses a time window sliding alignment algorithm to splice status data packets from different aggregation nodes and different interruption periods along the time axis, and marks the fault evolution timeline of each corresponding tower according to the fault determination metadata; for the sections with missing data on the time axis, the dual-modal reconstruction strategy is executed, and finally the continuous status trajectory and complete fault evolution time sequence of each monitoring point along the transmission line during the communication interruption are output.

10. A multi-dimensional sensing system for the status of transmission lines based on multi-source data fusion, characterized in that, The system includes: The tower-side data acquisition and caching unit includes multiple types of sensors, a unified timing module, and a local data caching module. It is used to collect multi-source heterogeneous status data and generate time stamps. In communication blind spots, it dynamically adjusts the caching strategy according to the link status. It also has a built-in lightweight edge inference unit for on-site determination of suspected fault status. The lightweight edge inference unit constructs a full-link value quantification model for transmission line fault association data in communication blind spots. The model is based on the fault association degree, time sensitivity, and physical importance of data packets as the basic dimensions, and superimposed with nonlinear gain of fault matching confidence, link communication state adaptation correction, and data retention reliability loss correction. The comprehensive value score V of the full-link data is obtained through a three-level progressive composite calculation. The local ad hoc network routing unit is deployed at the aggregation node and adjacent towers to build a hierarchical tree-like ad hoc network topology and to perform dynamic priority forwarding based on the tower status flag and the comprehensive value score V of the full-link data. The satellite backhaul and scheduling unit, deployed at the aggregation node, includes a satellite communication terminal and a data lifecycle hierarchical scheduling module, which is used to prioritize and hierarchically transmit data packets based on the comprehensive value score V of the entire link data within the satellite window period; The remote monitoring center includes a data receiving module, a time window sliding alignment module, a dual-modal reconstruction module, and a fault tracing module. It is used to splice data packets from different interruption periods, perform dual-modal reconstruction on missing data segments, and output continuous state trajectories and fault evolution sequences.

Images